01 · Compliance
18 Guardians with defined review responsibilities, designed to flag sensitive work for appropriate approval.
Guardians: 18 Compliance Reviewers
18 rolesBuilt · not yet in app
Safety & oversight
Five Essential Safeguards.
The prototype drafts work. Our protection framework is designed to support review.
Specialized oversight roles
Five connected layers describe the oversight architecture. A described responsibility is not a claim that every control is already enforced in code.
02 · Quality and Honesty
MAX is designed to provide independent evaluation and scoring. Evaluator independence and technical score-integrity enforcement are distinct; enforcement awaits verified implementation status.
MAXBuilt · not yet in app
03 · Security Review
Techno is designed for architecture, security and privacy reviews and technical plans. MARSHALL is the AI security Guardian.
TechnoBuilt or designed*MARSHALLBuilt · not yet in app
04 · Attack Testing
RED is designed to deliberately test the workforce through authorized challenges to help identify weaknesses before release.
REDBuilt · not yet in app
05 · Approval Rules
The human-control framework is designed to govern consequential actions such as sending, publishing, paying or committing. Technical approval enforcement awaits verified testing.
PlannedThe 18 Guardians
Eighteen specialized reviewers designed to stand between the team's work and your decisions, to review sensitive work, flag concerns and escalate decisions to people. Their enforcement controls are not yet verified as operational.
Review sensitive work
Each Guardian is designed with its own review responsibility, so sensitive work can be examined by the right reviewer.
Guardians: 18 Compliance Reviewers
18 rolesBuilt · not yet in app
Flag concerns
Designed to raise weak evidence, compliance questions and risky recommendations before a result reaches you.
Guardians: 18 Compliance Reviewers
18 rolesBuilt · not yet in app
Route to human approval
Designed so matters that need a decision go to a person. AI Team One is designed to prevent unauthorized external sends, financial transactions, trades, signatures and bookings; these safeguards are not yet verified as code-enforced.
PlannedWork alongside MAX, Techno and RED
The Guardians include MARSHALL for AI security. Independent evaluation, technology and security review, and red-team testing have separate roster roles.
MAXBuilt · not yet in appTechnoBuilt or designed*REDBuilt · not yet in app
Guardians: 18 Compliance Reviewers
18 rolesBuilt · not yet in app
The complete vision
The Customer Remains in Control.
Human approval checkpoints
Designed so people review consequential decisions. Proposed multi-step approvals are not presented as fully installed.
PlannedDraft-first execution
The prototype is read-and-draft only. Nothing is sent, posted, purchased or booked.
Source and evidence review
Designed to support review of evidence and important claims before a person relies on a result.
PlannedDefined agent permissions
Designed to define role boundaries and responsibilities, not unrestricted tool access.
PlannedIndependent grading
Evaluation is designed to be separate from authorship. Technical score protection remains subject to verification.
MAXBuilt · not yet in app
Emergency Shutdown (in preview)
Built so the owner can pause new tasks with one control. A step already running finishes before the pause takes effect. Testing is ongoing.
Available in previewLogging and review controls
Designed to support activity records and incident review within the intended control framework.
PlannedA prototype, not a certification claim
Testing and verification
AI Team One is in pre-launch development. No SOC 2, HIPAA, ISO certification or independently verified penetration-test result is claimed.
Synthetic test data only
Do not enter real business, financial, medical or client information into this early prototype.